Hans W. Chen;
Fuqing Zhang;
Thomas Lauvaux;
Richard B. Alley;
Kenneth J. Davis
The Pennsylvania State University
The 8th EnKF Data Assimilation Workshop
Constraining regional-scale CO2 fluxes using a coupled meteorological-carbon ensemble Kalman filter
Talk:
HansChen_EnKF_workshop_2018.pdf
Ensemble filter methods are becoming increasingly more popular for non-numerical weather prediction applications, for example to estimate sources and sinks of atmospheric CO2. Traditional inverse modeling techniques have shown promise in constraining CO2 fluxes using measurements of atmospheric CO2 concentration combined with an atmospheric transport model, but computational challenges arise for high resolution inversions and when the number of observations is high, for example for high-density remote sensing measurements. Furthermore, it is not straightforward how to represent the many different sources of uncertainties, including the uncertainty in the atmospheric transport. Data assimilation methods can potentially provide computationally efficient methods to solve the inverse problem, but there are several special considerations that have to be taken into account, such as the lack of a skillful forecast model for the CO2 fluxes, and the many sources of errors. Here we present a coupled meteorological-carbon ensemble Kalman Filter (EnKF) system for estimating CO2 fluxes at the regional scale. In this system, the meteorological state vector has been augmented to include atmospheric CO2 concentration and surface CO2 fluxes, which allows the EnKF to optimize both meteorological and carbon variables simultaneously. We show how errors in the simulated atmospheric transport can impact the estimated CO2 fluxes, and suggest methods to account for transport errors in the inversion system. Finally, we discuss some new data assimilation techniques, such as inflation and localization, that have been specifically adapted for the CO2 inverse problem.
Ensemble filter methods are becoming increasingly more popular for non-numerical weather prediction applications, for example to estimate sources and sinks of atmospheric CO2. Traditional inverse modeling techniques have shown promise in constraining CO2 fluxes using measurements of atmospheric CO2 concentration combined with an atmospheric transport model, but computational challenges arise for high resolution inversions and when the number of observations is high, for example for high-density remote sensing measurements. Furthermore, it is not straightforward how to represent the many different sources of uncertainties, including the uncertainty in the atmospheric transport. Data assimilation methods can potentially provide computationally efficient methods to solve the inverse problem, but there are several special considerations that have to be taken into account, such as the lack of a skillful forecast model for the CO2 fluxes, and the many sources of errors. Here we present a coupled meteorological-carbon ensemble Kalman Filter (EnKF) system for estimating CO2 fluxes at the regional scale. In this system, the meteorological state vector has been augmented to include atmospheric CO2 concentration and surface CO2 fluxes, which allows the EnKF to optimize both meteorological and carbon variables simultaneously. We show how errors in the simulated atmospheric transport can impact the estimated CO2 fluxes, and suggest methods to account for transport errors in the inversion system. Finally, we discuss some new data assimilation techniques, such as inflation and localization, that have been specifically adapted for the CO2 inverse problem.
Thanks to McGill's Atmospheric and Oceanic Sciences department for hosting this web-site.